Search for: All records

Abstract Private groundwater wells can be unmonitored sources of contaminated water that can harm human health. Developing models that predict exposure could allow residents to take action to reduce risk. Machine learning models have been successful in predicting nitrate contamination using geospatial information such as proximity to nitrate sources, but previous models have not considered meteorological factors that change temporally. In this study, we test random forest (regression and classification) and linear regression models to predict nitrate contamination using rainfall, temperature, and readily available soil parameters. We trained and tested models for (1) all of North Carolina, (2) each geographic region in North Carolina, (3) a three‐county region with a high density of animal agriculture, and (4) a three‐county region with a low density of animal agriculture. All regression models had poor predictive performance (R² < 0.09). The random forest classification model for the coastal plain showed fair agreement (Cohen'sκ = 0.23) when trying to predict whether contamination occurred. All other classification models had slight or poor predictive performance. Our results show that temporal changes in rainfall and temperature, or in combination with soil data, are not enough to predict nitrate contamination in most areas of North Carolina. The low level of contamination (<25%) measured during the study could have contributed to the poor performance of the models. 

Abstract Early forecasts give people in a storm’s path time to prepare. Less is known about the cost to society when forecasts are incorrect. In this observational study, we examine over 700,000 births in the path of Hurricane Irene and find exposure was associated with impaired birth outcomes. Additional warning time was associated with decreased preterm birth rates for women who experienced intense storm exposures documenting a benefit of avoiding a type II forecasting error. A larger share of this at-risk population experienced a type I forecasting error where severe physical storm impacts were anticipated but not experienced. Disaster anticipation disrupted healthcare services by delaying and canceling prenatal care, which may contribute to storm-impacted birth outcomes. Recognizing storm damages depend on human responses to predicted storm paths is critical to supporting the next generation’s developmental potential with judicious forecasts that ensure public warning systems mitigate rather than exacerbate climate damages. 

Domestic wells provide drinking water to 44 million people nationwide. Many of these wells, which remain federally unregulated and rarely tested for pollutants, serve rural populations clustered near surface-contaminated sites (e.g., hazardous waste sites, animal agriculture operations, coal ash ponds, etc.). The potential for natural disasters to deteriorate drinking water quality is well documented. Less understood is whether opportunistic post-disaster sampling might underrepresent vulnerable populations. When disaster strikes, well water sampling campaigns offer a glimpse into the quality of water for exposed residents. We examined over 8,000 well water samples from 2016 and 2017 to measure Hurricane Matthew’s impact on the presence of indicator bacteria. Bacteria presence was predicted at the household level following Hurricane Matthew’s landfall. The residential addresses associated with birth records as well as clinically estimated dates of conception and birth dates were used to predict the likelihood of indicator bacteria in drinking water sources that were unsampled but likely to have served pregnant women. We estimate that opportunistic well water sampling captures the average predicted contamination rates among households with pregnant women. Our approach documents a distribution of contamination risk where 2.7% of the vulnerable sample (670 unsampled households) have a 75% likelihood of total coliform presence. The predicted likelihood of indicator bacteria is elevated for a small share of households nearby swine lagoons that experienced the most torrential rainfall. However, the gap between sampled and unsampled households cannot otherwise be explained by the storm event or proximity to surface-contaminated sites. Findings suggest that sophisticated and holistic water quality prediction models may support post-disaster sampling campaigns by targeting individual households within vulnerable groups that are likely to experience higher risks from groundwater contamination.